[CGP] don't expand a memcmp with nobuiltin attribute

[spatel]

Summary

This matches the behavior used in the SDAG when expanding memcmp.

For reference, we're intentionally treating the earlier fortified call transforms differently after:
https://bugs.llvm.org/show_bug.cgi?id=23093
https://reviews.llvm.org/rL233776

One motivation for not transforming nobuiltin calls is that it can interfere with sanitizers:
https://reviews.llvm.org/D19781
https://reviews.llvm.org/D19801

[efriedma]

LGTM with one minor tweak.

include/llvm/Analysis/TargetLibraryInfo.h

246 ↗

(On Diff #101949)

This should check that CS.getCalledFunction() is nonnull. (CodeGenPrepare::optimizeCallInst checks this explicitly, but other code might not.)

llvm/trunk/include/llvm/Analysis/TargetLibraryInfo.h

 //===-- TargetLibraryInfo.h - Library information ---------------*- C++ -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_ANALYSIS_TARGETLIBRARYINFO_H
 #define LLVM_ANALYSIS_TARGETLIBRARYINFO_H
 
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/Optional.h"
 #include "llvm/ADT/Triple.h"
+#include "llvm/IR/CallSite.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/PassManager.h"
 #include "llvm/Pass.h"
 
 namespace llvm {
 template <typename T> class ArrayRef;
 
public:
   bool getLibFunc(StringRef funcName, LibFunc &F) const {
     return Impl->getLibFunc(funcName, F);
   }
 
   bool getLibFunc(const Function &FDecl, LibFunc &F) const {
     return Impl->getLibFunc(FDecl, F);
   }
 
+  /// If a callsite does not have the 'nobuiltin' attribute, return if the
+  /// called function is a known library function and set F to that function.
+  bool getLibFunc(ImmutableCallSite CS, LibFunc & F) const {
+    return !CS.isNoBuiltin() && CS.getCalledFunction() &&
+           getLibFunc(*(CS.getCalledFunction()), F);
+  }
+
   /// Tests whether a library function is available.
   bool has(LibFunc F) const {
     return Impl->getState(F) != TargetLibraryInfoImpl::Unavailable;
   }
   bool isFunctionVectorizable(StringRef F, unsigned VF) const {
     return Impl->isFunctionVectorizable(F, VF);
   }
   bool isFunctionVectorizable(StringRef F) const {

llvm/trunk/lib/CodeGen/CodeGenPrepare.cpp

bool CodeGenPrepare::optimizeCallInst(CallInst *CI, bool& ModifiedDT) {
   FortifiedLibCallSimplifier Simplifier(TLInfo, true);
   if (Value *V = Simplifier.optimizeCall(CI)) {
     CI->replaceAllUsesWith(V);
     CI->eraseFromParent();
     return true;
   }
 
   LibFunc Func;
-  if (TLInfo->getLibFunc(*CI->getCalledFunction(), Func) &&
-      Func == LibFunc_memcmp) {
-    if (expandMemCmp(CI, TTI, TLI, DL)) {
-      ModifiedDT = true;
-      return true;
-    }
-  }
+  if (TLInfo->getLibFunc(ImmutableCallSite(CI), Func) &&
+      Func == LibFunc_memcmp && expandMemCmp(CI, TTI, TLI, DL)) {
   return false;
 }
 
 /// Look for opportunities to duplicate return instructions to the predecessor
 /// to enable tail call optimizations. The case it is currently looking for is:
 /// @code
 /// bb0:
 ///   %tmp0 = tail call i32 @f0()

llvm/trunk/test/CodeGen/PowerPC/memCmpUsedInZeroEqualityComparison.ll

; CHECK-NEXT:    blr
   %not.tobool = icmp eq i32 %call, 0
   %cond = zext i1 %not.tobool to i32
   ret i32 %cond
 }
 
 define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) {
 ; CHECK-LABEL: length2_eq_nobuiltin_attr:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    lhz 3, 0(3)
-; CHECK-NEXT:    lhz 4, 0(4)
-; CHECK-NEXT:    li 5, 0
-; CHECK-NEXT:    li 12, 1
-; CHECK-NEXT:    cmpw 3, 4
-; CHECK-NEXT:    isel 3, 12, 5, 2
+; CHECK-NEXT:    mflr 0
+; CHECK-NEXT:    std 0, 16(1)
+; CHECK-NEXT:    stdu 1, -32(1)
+; CHECK-NEXT:  .Lcfi0:
+; CHECK-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NEXT:  .Lcfi1:
+; CHECK-NEXT:    .cfi_offset lr, 16
+; CHECK-NEXT:    li 5, 2
+; CHECK-NEXT:    bl memcmp
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    cntlzw 3, 3
+; CHECK-NEXT:    rlwinm 3, 3, 27, 31, 31
+; CHECK-NEXT:    addi 1, 1, 32
+; CHECK-NEXT:    ld 0, 16(1)
+; CHECK-NEXT:    mtlr 0
 ; CHECK-NEXT:    blr
   %m = tail call signext i32 @memcmp(i8* %X, i8* %Y, i64 2) nobuiltin
   %c = icmp eq i32 %m, 0
   ret i1 %c
 }